Industry 4.0 Adoption Using AI/ML-Driven Metamodels for High-Performance Ductile Iron Sand Casting Design and Manufacturing

Abstract

Data-centric near-real-time intelligent process control for smart manufacturing in an Industry 4.0 era is of tremendous value. Design and manufacturing of high-performance ductile iron sand castings is a multi-variant complex process with much uncertainty involved. As a result, in spite of a well-controlled operation and an experienced workforce, iron foundries in a production environment do face sporadic shrinkage and lots with nonconforming property requirements, resulting in scrap or rework. A framework and methodology consisting of AI (artificial intelligence) and ML (machine learning) tools, coupled with ICME (integrated computational materials engineering) and process simulation tools, will be presented to quantify uncertainty (UQ). Metamodels, both predictive and prescriptive in near real time were developed using such AI/ML techniques using historical production and selective design of experiments (DOE)-generated additional data. The data will be presented including details on successful corrective action production trials. The proposed framework and approach is applicable to solve such complex problems encountered in the foundry and machining operations where there is uncertainty.